Neonatal Lung Sonography: Interobserver Agreement Between Physician Interpreters With Varying Levels of Experience

Objectives: To assess the reliability of lung sonography in neonates between physician interpreters with different degrees of experience.

Methods: We retrospectively reviewed lung sonograms from neonates admitted to a neonatal intensive care unit with respiratory distress in the first 24 hours of life. The first scans were selected; only patients with available video clips documenting both hemithoraxes were included. The clips were independently examined by 4 different experienced observers blinded to clinical data. The interpreting physicians made a codified sonographic diagnosis, and the Cohen κ coefficient was used to measure the reliability between a proven experienced main interpreter and expert (κ1), intermediate (κ2), and beginner (κ3) control interpreters. We also calculated the specific agreement on respiratory distress syndrome and transient tachypnea of the neonate.

Results: Four hundred sixty-five clips were taken from 114 neonates examined over a 16-month period. The patients' median gestational age (range) was 34 weeks (25-41 weeks), and the median birth weight (range) was 2085 g (608-4134 g). Eighty-eight percent of examinations were performed within 24 hours after birth. The overall κ coefficients (95% confidence intervals) were κ1 = 0.94 (0.88-1.00); κ2 = 0.72 (0.61-0.83); and κ3 = 0.81 (0.71-0.90). For respiratory distress syndrome, κ1 = 0.94 (0.87-1.00); κ2 = 0.90 (0.81-0.99); and κ3 = 0.87 (0.78-0.97). For transient tachypnea of the neonate, κ1 = 0.95 (0.89-1.00); κ2 = 0.76 (0.64-0.88); and κ3 = 0.81 (0.70-0.91).

Conclusions: In neonates with early respiratory distress, lung sonography has high interobserver agreement even between interpreters with varying levels of experience.